{
\def\sym#1{\ifmmode^{#1}\else\(^{#1}\)\fi}
\begin{tabular}{l*{6}{c}}
\toprule
            &\multicolumn{2}{c}{\specialcell{Contemporaneous\\Effect (t)}}&\multicolumn{2}{c}{\specialcell{One Year\\Later (t+1)}}&\multicolumn{2}{c}{\specialcell{Previous Year\\(Placebo) (t-1)}}\\\cmidrule(lr){2-3}\cmidrule(lr){4-5}\cmidrule(lr){6-7}
            &\multicolumn{1}{c}{\specialcell{Number\\(1)}}&\multicolumn{1}{c}{\specialcell{IHS\\(2)}}&\multicolumn{1}{c}{\specialcell{Number\\(3)}}&\multicolumn{1}{c}{\specialcell{IHS\\(4)}}&\multicolumn{1}{c}{\specialcell{Number\\(5)}}&\multicolumn{1}{c}{\specialcell{IHS\\(6)}}\\
\midrule
Neighbor $\times$ Corner & -.07 & .0015 & .29 & .089 & -.14 & .038 \\ & [.61] & [.98] & [.075]* & [.12] & [.48] & [.6] \\ & \{.67\} & \{.98\} & \{.15\} & \{.17\} & \{.45\} & \{.56\} \\ & <1> & <1> & <.92> & <.92> & & \\ \\[-5pt] Neighbor $\times$ Middle & -.0085 & -.0071 & .043 & .011  & -.0051 & .0055 \\  & [.92] & [.8] & [.71] & [.78] & [.95] & [.89] \\ & \{.9\} & \{.76\} & \{.69\} & \{.76\} & \{.95\} & \{.88\} \\ & <1> & <1> & <1> & <1> & & \\ \\[-5pt] Corner = Middle & [.62] & [.87] & [.071]* & [.16] & [.55] & [.7] \\ & \{.7\} & \{.87\} & \{.3\} & \{.3\} & \{.51\} & \{.67\} \\ \\[-5pt] \hline Observations & 22687 & 22687 & 15172 & 15172 & 15130 & 15130 \\ Session $\times$ Corner FE & Y & Y & Y & Y & Y & Y \\ Session $\times$ Party Pair FE & Y & Y & Y & Y & Y & Y \\ Outcome Mean & 1.9 & .98 & 1.9 & .97 & 1.8 & .93 \\ Outcome S.d. & 2.7 & .94 & 3 & .97 & 2.6 & .92 \\ \toprule \multicolumn{7}{p{14.5cm}}{\footnotesize \textit{Notes:} Regressions include different-party dyads only, with neither MP pre-assigned seats. Number is the total number of co-sponsorship links between the two MPs in a pair in a given session. IHS is the inverse hyperbolic sine transformation of Number. Neighbor is a dummy variable equal to one if the MPs in the pair are randomly assigned to sit next to each other during that session. Corner is equal to one if at least one MP in pair has only one seating neighbor. Middle is equal to one minus Corner. Dyadic-robust p-values are in square brackets. Randomization inference p-values (1000 draws) are in curly brackets. Sharpened q-values (Anderson 2008) for non-placebo tests are in <>. Special sessions and a short session (2017) are excluded. For lead and lag specifications, sessions are also dropped where lead/lag would be a special/short session or a session in a different parliamentary term. *** p<0.01, ** p<0.05, * p<0.1.} \\ \end{tabular} }
